Run the combined benchmark before the 4.2.0 release #1288

Summary

Run one combined benchmark on the commit planned for 4.2.0, and release 4.2.0 on that commit unless the results show a major problem. Collecting the measurements in one run keeps the cost and effort to a single run, and ties the published results to the released commit.

Sub-issues

• Benchmark the temporal type conversion changes of 4.2.0 #1287: the temporal type conversion changes, measured on 4.2.0 only (the PRs' local measurements give the comparison). It needs a temporal shape in the harness, merged before the commit planned for 4.2.0.
• Benchmark chunked reads of a result larger than host memory #1162: chunked reads of a result larger than host memory (100 million rows).
• Size the asyncio filesystem's parallel requests by max_workers instead of the default executor #1159: the asyncio filesystem's parallel requests, sized by max_workers. The change is merged, and this issue stays open for the measurement.

Plan

• Use Spot hosts, on the same day:
  • an r7g.2xlarge stack for the cursor and read_sql_query() cases, including Benchmark the temporal type conversion changes of 4.2.0 #1287 and Benchmark chunked reads of a result larger than host memory #1162;
  • a c8gn.2xlarge stack for the filesystem suite (Size the asyncio filesystem's parallel requests by max_workers instead of the default executor #1159).
• Measure the commit planned for 4.2.0. Unless the results show a major problem, release 4.2.0 on that commit. Improvements found in the results are filed as issues and made after the release.
• After the run:
  • Decide the AioS3FileSystem default for writes and copies (Size the asyncio filesystem's parallel requests by max_workers instead of the default executor #1159, following Benchmark S3ThreadPoolExecutor vs S3AioExecutor before defaulting to async #685).
  • If the results hold, close larger than memory queries? #61 (through Benchmark chunked reads of a result larger than host memory #1162) and SQLAlchemy + Pandas very slow when compared to AWS Wrangler #254.
  • Record the results under benchmarks/history/ and in benchmarks/RESULTS.md.